基于高斯模型的不确定数据流Skyline查询方法

由于数据的动态性及不确定性等特征,使得不确定数据流上Skyline查询研究面临挑战.不确定对象一般采用多元概率密度函数(PDF)表示,现有的不确定数据流Skyline查询方法均采用离散型随机变量建模.然而不确定数据流中的对象可能是连续变化的,离散模型对连续性随机变量难以适用.针对连续PDF建模的不确定数据流Skyline查询进行了研究,提出了基于高斯模型的不确定数据流Skyline查询方法(SGMU),该方法包含2个过程:1)动态高斯建模算法(DGM):对滑动窗口采样并建立高斯模型,将原始的数据流转化为不确定对象PDF的参数流;2)提出了基于高斯树的查询算法(GTS)以建立空间索引结构和执行Skyline查询.实验结果表明,SGMU算法不仅能够对连续型不确定对象进行有效建模以辅助Skyline查询,而且能够有效地减少查询对象个数,提高Skyline查询效率.

A Skyline Query Method over Gaussian Model Uncertain Data Streams

Skyline queries over uncertain streams has become a challenge because of the uncertainty and dynamics of data. Uncertain data is usually represented by a multivariate probability density function (PDF). The current skyline queries over uncertain data streams are modeled by discrete PDF. However, the discrete PDF may not suit for many streams, because these streams may be continuously changing. The research of the skyline query on continuous PDF model proposes an efficient skyline query method over Gaussian model uncertain stream (SGMU). Firstly, a dynamic Gaussian modeling (DGM) algorithm is proposed to build the Gaussian model by sampling sliding window in streams. Secondly, a Gauss-tree based skyline query algorithm (GTS) is raised to build a spatial indexing structure for uncertain data stream. Experimental results demonstrate that SGMU not only can model the uncertain objects efficiently to support skyline queries, but also can greatly improve the skyline queries by pruning objects efficiently.